Effort levels by model › Claude Haiku 5.5
Claude Haiku 5.5 reasoning effort levels: intelligence, cost and speed
How much smarter Claude Haiku 5.5 gets at each reasoning effort, what each level costs per task, and how fast it answers. Benchmarks from Artificial Analysis.
| Effort | Intelligence Index | Gain vs. previous | Cost per task | Output tokens/s |
|---|---|---|---|---|
| low Best intelligence per dollar | 19 | — | $0.04 | 168 |
| medium | 25 | +6 | $0.09 | 176 |
| high | 30 | +5 | $0.16 | 182 |
| xhigh | 34 | +4 | $0.25 | 200 |
| max | 36 | +2 | $0.41 | 236 |
API: $0.1 / $0.5 per 1M input / output tokens.
Source: Artificial Analysis LLM Leaderboard, checked 2026-11-02.